AddThis offers online tools that leverage data from our network of 15 million sites. We help publishers understand their audiences and drive traffic to what matters.



Behind the Scenes of Our Scrolling Engagement Report

This morning, we released our 2014 Q2 Engagement Report analyzing scrolling behavior on content across the AddThis network. In this report, we break this data down broadly by time and operating system, but also go deeper into how users were referred to the page (i.e. through ad campaigns), and which AddThis tools the pages were using. Here I’ll describe the mechanics of how we created the report.

To collect this data, little was changed about the way AddThis tools operate on publisher sites. Our tools take scrolling into account already. For example, our “What’s Next” layer appears when users are about 50% of the way through page content in an attempt to recirculate a user to another article that interests them. At very specific scrolling intervals, we send signals that scrolling occurred to our servers in order to measure the tool’s visibility, and help in reporting and insights for our customers.

The scrolling data, like all data we collect, is tied to the browser ID, which is a randomly-generated cookie we set in all browsers that have not opted-out of AddThis personalization. However, even if a given user did opt-out, we collect this data associated with the URL only––not the browser. This allows us to continue to optimize and report on scrolling for our publishers.

Now that we have this anonymous data, we use our open-sourced data processing system Hydra to ask questions about it. Hydra is great for summarizing huge datasets with large cardinalities (number of individual elements in the set of data), which is what we deal with most of the time. To draw our conclusions, we looked at dominating statistics for billions of users, which yielded statistically-significant answers, and considered only large groups of behaviors without focusing on small sets of browser data.

You can download the full report and take a closer look at the findings. We love understanding behavior like this, because it helps our publishers’ sites perform better while being more relevant to the consumers viewing them. Feel free to reach out in the comments with any questions.

  • Interesting stats, wonder what it means for me as an iOS user..

    One of the findings was the bounce rate from social referrals. I am curious to know how many of the bounce traffic, might save the article for later. I find that many times I’ll quickly skim an article and if I find it interesting I’ll save it for later, how does that figure in to the findings?

  • Rich L

    Hey Ben. Good point on saving for later.

    It’s hard to pick out those events unless they are triggered through AddThis (we actually have all the bookmarking / read later services in our menu). If they copy the URL and then exit Safari or Chrome and let the native app pick up the URL to add, we could detect the copy but not the resulting action.

    I would like to do a study to see if we get referrals back to content from the read later apps/services, but the native apps like Pocket may not present a clear referring URL to detect. Would be an interesting project to see how much we can detect though – I’m sure pubs would love to know.